Confidence score
A number a model or classifier reports to indicate how sure it is. It is only useful for decisions if it is calibrated, and it never decides on its own what action is permitted.
A confidence score is the number a system attaches to an answer to express how likely it believes the answer is right. Confidence can come from a probability output, a classifier margin or a separate estimator.
Scores are easy to misuse. A high number is not proof, and the same number can mean different things in different systems. They are trustworthy only after calibration: checking, on your own traffic, that cases scored at 80% are right roughly 80% of the time.
Confidence is not permission
Even a well-calibrated score answers only "what is likely to be true?". It does not answer "what is this system allowed to do if it is true?". That second question belongs to a threshold policy and to authority rules owned by people. Low scores should route somewhere real, which is abstention.
Read more in OpenAI Decisions API turns probability into policy.
Related terms
Calibration
How well a model's stated confidence matches reality: if it says 90% on a hundred comparable cases, roughly ninety should be correct. Once probabilities drive actions, it becomes an operating metric.
Threshold policy
The explicit rules that turn a model's probability or score into an action — reject, send to a human, or act autonomously — owned and reviewed separately from the model itself.
Abstention
A deliberate option for an AI decision system to decline to decide and return uncertain or high-consequence cases to a person, rather than forcing every input into yes or no.